Multi-objective optimized allocation of arid saline farmlands and irrigation water resources for sustainable agriculture

Agricultural production in arid regions faces challenges from water scarcity, soil salinization, and land and water resources allocation. To address these issues, a multi-objective framework for the allocation of saline farmlands and irrigation water resources was formulated, incorporating rarely-considered root-zone salinity dynamics. Taking the arid oasis Manas River Basin in Xinjiang, China as an example, a verified remote sensing inversion method was applied to estimate the spatial distribution of 0–80 cm root-zone soil salt content ( SSC ) for different crops. This was then used to train XGBoost models to predict root-zone SSC , total actual evapotranspiration, and annual maximum NDVI under various land and water allocation scenarios. Based on regional goals of reducing irrigation from 1.58 × 10 9 to 1.20 × 10 9 m 3 , NSGA-II was employed to optimize land and water allocation with three objectives including maximum total cotton yield ( TCY ), minimum SSC , and maximum irrigation water productivity ( WP I ). The XGBoost models achieved high accuracy ( R 2 > 0.90). Optimization showed root-zone SSC could fall by 0.06 g kg −1 , with minimal WP I improvement , and TCY drops by over 2 × 10 8 kg, representing ∼28 % reduction from the 2022 baseline of 7.08 × 10⁸ kg, due to reduced planting area. Referring to SSC inversion data, crop area was adjusted in GIS by removing higher-salinity pixels (ranked in descending order), with an average area deviation of less than 0.8 % from optimization results. By integrating high-resolution SSC estimation, XGBoost and NSGA-II algorithm, this study offers a practical framework for precise saline farmland management and efficient water use in arid regions. • Root-zone salinity control is integrated into multi-objective optimization. • XGBoost enables precise prediction of root-zone salinity from land and water use. • Land-water allocation is regulated by combining XGBoost and NSGA-II. • Multi-objective optimization allows adjustment of crop planting structure.